Wind Power Prediction Based on the Clustered Combination of ARMA and PSO-SVM Methods

Abstract

With the increase of installed capacity of wind power, the fluctuation of wind power has more and more obvious impact on the stable operation of power system. Wind power prediction is the key technology to improve the stability level of power system with large-scale integration of wind power. Based on the historical data of wind power, wind speed and temperature, this paper establishes an autoregressive moving average (ARMA) prediction model and a support vector machine (SVM) prediction model, moreover, Particle swarm optimization (PSO) algorithm is used to solve the problem of parameter optimization of SVM model. Considering the wind power prediction results of the two models, a combined prediction model based on ARMA model and PSO-SVM model is established based on covariance minimization method, and with the basis of clustering theory, a C-PSO-SVM-ARMA clustered combination model is proposed. In case study, prediction performance of different models is examined, and the clustered combination model C-PSO-SVM-ARMA model is verified as the best wind power prediction model.